Instruction Following & Long Context

Artificial Analysis public leaderboard

IFBench

Instruction-following benchmark results published by Artificial Analysis.

Published source

Metric

Accuracy

Results shown

450

Publisher

Artificial Analysis

Snapshot

Fetched 2026-08-22T12:55:16.136Z

450 of 450 models

Higher is better

Grok 4.3 (medium)

Grok 4.20 0309

MiniMax-M3

Nemotron 3 Ultra

Grok 4.3 (high)

Grok 4.20 0309 v2

Grok 4.3 (low)

Qwen3.7 Max

Nemotron Cascade 2 30B A3B

MiMo-V2.5-Pro

Nova 2.0 Pro Preview (low)

DeepSeek V4 Flash (max)

Nova 2.0 Pro Preview (medium)

Qwen3.5 397B A17B

Gemini 3 Flash

Qwen3.7 Plus

GPT-5.2 Codex (xhigh)

Gemini 3.1 Flash-Lite

Gemini 3.1 Pro Preview

Qwen3.6 Max Preview

DeepSeek V4 Pro (max)

Gemini 3.5 Flash

GLM-5.1

Kimi K2.6

GPT-5.4 nano (xhigh)

Muse Spark

GPT-5.5 (xhigh)

Qwen3.5 122B A10B

MiniMax-M2.7

Qwen3.5 27B

Gemma 4 31B

GPT-5 mini (high)

GPT-5.2 (xhigh)

GPT-5.3 Codex (xhigh)

Qwen3.6 Plus

Gemini 3.5 Flash (medium)

GPT-5 Codex (high)

Command A+

GPT-5.4 (xhigh)

Gemma 4 12B

DeepSeek V4 Flash (high)

GLM-5.2 (max)

GPT-5.4 mini (xhigh)

GLM-5-Turbo

GPT-5 (high)

GPT-5.1 (high)

GPT-5.6 Sol (max)

Qwen3.5 35B A3B

Gemma 4 26B A4B

GLM-5

MiniMax-M2

MiMo-V2-Flash (Feb 2026)

MiniMax-M2.5

GPT-5.5 (high)

Nemotron 3 Super

GPT-5.5 Instant (May 2026)

DeepSeek V4 Pro (high)

GPT-5.6 Terra (max)

Solar Pro 3

GPT-5 mini (medium)

Nemotron 3 Nano

GPT-5.6 Sol (xhigh)

GPT-5.5 (medium)

Nova 2.0 Lite (high)

Qwen3 Max Thinking

GPT-5 (medium)

Gemini 3 Pro Preview (high)

Kimi K2.5

GPT-5.1 Codex (high)

MiniMax-M2.1

Mercury 2

GPT-5.6 Sol (medium)

GPT-5.6 Sol (high)

Apriel-v1.6-15B-Thinker

gpt-oss-120b (high)

MiMo-V2-Pro

Mistral Medium 3.5

o4-mini (high)

Nova 2.0 Lite (medium)

KAT-Coder-Pro V1

Kimi K2 Thinking

GLM-4.7

GPT-5.1 Codex mini (high)

Qwen3.6 27B

GPT-5 nano (high)

MiMo-V2-Omni-0327

Step 3.7 Flash

o3-mini (high)

MiMo-V2.5

Qwen3.5 9B

KAT-Coder-Pro V2

GPT-5 (low)

GPT-5.6 Sol (low)

Step 3.5 Flash 2603

HyperNova 60B 2605

GPT-5.6 Terra (xhigh)

Nova 2.0 Omni (medium)

Nex-N2-Pro

Olmo 3.1 32B Think

GPT-5.4 (low)

GPT-5 nano (medium)

GPT-5.2 (medium)

gpt-oss-20b (high)

GPT-5.4 mini (medium)

K-

K-EXAONE

Step 3.5 Flash

GPT-5.6 Terra (high)

GPT-5.4 nano

Qwen3.6 35B A3B

GPT-5.5 (low)

MiMo-V2-Flash

DeepSeek V3.2 Speciale

Claude Fable 5 (with fallback)

Nemotron 3 Nano Omni 30B A3B

Hy3-preview

Kimi K2.7 Code

K2 Think V2

Claude Opus 4.8 (max)

GPT-5.6 Terra (medium)

Nova 2.0 Omni (low)

Apriel-v1.5-15B-Thinker

Nova 2.0 Lite (low)

GLM 5V Turbo

GLM-4.7-Flash

Qwen3 Next 80B A3B

DeepSeek V3.2

K2-V2 (high)

GPT-5.6 Terra (low)

DiffusionGemma 26B A4B

Qwen3 VL 32B

Claude Opus 4.7 (max)

gpt-oss-120b (low)

Nemotron 3 Nano 4B

Claude Opus 4.5

EXAONE 4.5 33B

gpt-oss-20b (low)

Solar Open 100B

North Mini Code

Ling 2.6 Flash

Claude 4.5 Sonnet

DeepSeek V3.1 Terminus

Motif-2-12.7B

Ling-2.6-1T

Claude Sonnet 4.6 (max)

Qwen3 VL 235B A22B

Trinity Large Thinking

LFM2.5-8B-A1B

Claude 4.1 Opus

GLM-5

Gemini 3 Flash

K2-V2 (medium)

Claude 4 Sonnet

Tri-21B-Think

GLM-4.7

Falcon-H1R-7B

Claude 4.5 Haiku

DeepSeek V3.2 Exp

Qwen3 Max Thinking (Preview)

Claude 4 Opus

Grok 4

MiMo-V2-Omni

Gemma 4 31B

Claude Opus 4.6 (max)

Grok 4.1 Fast

Gemini 2.5 Flash-Lite (Sep)

Jamba Reasoning 3B

Gemini 2.5 Flash (Sep)

Nova 2.0 Pro Preview

Qwen3.5 4B

GLM-5.1

Qwen3.5 397B A17B

Doubao Seed Code

Qwen3 235B A22B 2507

Qwen3.5 Omni Plus

Qwen3.5 122B A10B

Qwen3 30B A3B 2507

Grok 4 Fast

Gemini 2.5 Flash

Step3 VL 10B

Gemini 2.5 Flash-Lite

Qwen3 4B 2507

Gemini 3 Pro Preview (low)

Grok 4.20 0309 v2

MiniCPM5-1B

Mi:dm K 2.5 Pro

Olmo 3 32B Think

DeepSeek V3.2

Gemini 2.5 Pro

GPT-5.4 (Non-reasoning)

Claude 3.7 Sonnet

Mistral Small 4

Qwen3 Max (Preview)

Hy3-preview

Grok 4.20 0309

Grok 4.3 (Non-reasoning)

GPT-5.2

Gemini 3.5 Flash (minimal)

DeepSeek V4 Flash

Tri-21B-think Preview

Llama 3.3 70B

Grok 3

Qwen3.5 27B

GLM-4.7-Flash

Cogito v2.1

GPT-5.5 (Non-reasoning)

Qwen3 235B 2507

LFM2 24B A2B

Grok 3 mini Reasoning (high)

DeepSeek V4 Pro

Qwen3.6 27B

GPT-5 mini (minimal)

GPT-5 (minimal)

Mi:dm K 2.5 Pro Preview

Gemma 4 26B A4B

Claude 4 Sonnet

Gemma 4 12B (Non-reasoning)

Qwen3 VL 30B A3B

GPT-5 (ChatGPT)

Claude Opus 4.6 (high)

Ring-1T

Ring-2.6-1T

Qwen3.5 35B A3B

Magistral Small 1.2

Granite 4.1 30B

Kimi K2.6

Gemma 4 E4B

Qwen3 Max

GLM-4.5

Claude 3.7 Sonnet

LFM2.5-1.2B-Instruct

Kimi K2.5

Claude Opus 4.7 (Non-reasoning, high)

Gemini 2.5 Flash (Sep)

GLM-4.6

Qwen3 Omni 30B A3B

Ring-flash-2.0

Claude 4 Opus

GPT-5.1

DeepSeek V3.2 Exp

LongCat Flash Lite

GPT-4.1

Claude Opus 4.5

Magistral Medium 1.2

Llama 4 Maverick

Claude 3.5 Haiku

MiMo-V2.5-Pro

Claude 4.5 Sonnet

Qwen3 VL 235B A22B

Claude Sonnet 4.6 (Non-reasoning, Low Effort)

JT-35B-Flash

Claude 4.5 Haiku

Seed-OSS-36B-Instruct

LFM2.5-1.2B-Thinking

Gemini 2.5 Flash-Lite (Sep)

MiniMax M1 80k

Kimi K2 0905

DeepSeek V3.1

Olmo 3 7B Think

Qwen3 30B

Kimi K2

Grok Code Fast 1

ERNIE 5.0 Thinking Preview

MiniMax M1 40k

DeepSeek V3.1 Terminus

Claude Sonnet 4.6 (Non-reasoning)

Nova 2.0 Omni

DeepSeek V3 0324

K2-V2 (low)

Qwen3 14B

Gemma 4 E4B

Nova 2.0 Lite

Qwen3 Coder 480B

Gemini 2.0 Flash

MiMo-V2-Flash

Qwen3 VL 8B

Mistral Medium 3.1

Qwen3 Next 80B A3B

K-

K-EXAONE

DeepSeek R1 0528

Llama 4 Scout

Llama 3.3 Nemotron Super 49B

Mistral Medium 3

Qwen3 VL 32B

Olmo 3.1 32B Instruct

ERNIE 4.5 300B A47B

Llama 3.1 405B

DeepSeek R1 (Jan)

Gemini 2.5 Flash

GPT-5.4 mini

QwQ-32B

Qwen3 235B

Granite 4.1 8B

GPT-4.1 mini

Llama Nemotron Ultra

Devstral 2

Llama 3.3 Nemotron Super 49B

OLMo 2 32B

Nova Pro

Gemma 4 E2B

Qwen3.5 Omni Flash

HyperCLOVA X SEED Think (32B)

Qwen3.5 9B

DeepSeek V3.1

Grok 4 Fast

GLM-4.5-Air

Nemotron 3 Nano

Llama 3 70B

Solar Pro 2

Llama Nemotron Super 49B v1.5

Qwen2.5 72B

Gemma 3 12B

GLM-4.6

JT-MINI

Qwen3 235B

Qwen3 VL 4B

Command A

Grok 4.1 Fast

EXAONE 4.0 32B

Qwen3 32B

Mistral Large 3

Qwen3.6 35B A3B

Nova Premier

Claude 3 Haiku

GPT-4o (Aug)

Nanbeige4.1-3B

Qwen3 Coder Next

Jamba 1.7 Large

MiniCPM5-1B

Ling-1T

DeepSeek V3 (Dec)

Hermes 4 405B

Devstral Small

Pixtral Large

Llama 3.1 70B

Sarvam 105B (high)

Ling-flash-2.0

GPT-4o (Nov)

GLM-4.5V

Nova Lite

INTELLECT-3

Solar Pro 2

Granite 4.1 3B

Gemma 4 E2B

Qwen3 4B 2507

EXAONE 4.0 32B

Mistral Small 3.2

Qwen3 8B

Qwen3.5 4B

GPT-4

LFM2.5-VL-1.6B

Qwen3 VL 30B A3B

Qwen3 30B A3B 2507

Llama Nemotron Super 49B v1.5

Mistral Small 4

Olmo 3 7B

Hermes 4 405B

GPT-5.4 nano

Qwen3 Coder 30B A3B

GPT-5 nano (minimal)

Qwen3 4B

Qwen3 VL 8B

Ministral 3 14B

GPT-4.1 nano

NVIDIA Nemotron Nano 12B v2 VL

Qwen3 30B

Gemma 3 27B

Sarvam M

Qwen3 VL 4B

Mistral Large 2 (Jul)

Devstral Small (May)

Gemini 2.5 Flash-Lite

Qwen3.5 2B

Granite 4.0 H Small

Qwen3 32B

Jamba 1.7 Mini

Hermes 4 70B

Mistral Large 2 (Nov)

Qwen3 Omni 30B A3B

Devstral Small 2

GPT-4o mini

Llama 3.1 Nemotron 70B

Llama 3.2 11B (Vision)

Reka Flash 3

GLM-4.6V

Mistral Small 3.1

Devstral Medium

Nova Micro

Qwen3.5 2B

Ministral 3 8B

Hermes 4 70B

GLM-4.5V

Llama 3.1 8B

Qwen3 8B

Gemma 3 4B

Kimi Linear 48B A3B Instruct

GLM-4.6V

Gemma 3n E4B

NVIDIA Nemotron Nano 9B V2

DeepSeek R1 Distill Llama 70B

NVIDIA Nemotron Nano 9B V2

Molmo2-8B

Qwen3 1.7B

Ministral 3 3B

MiniCPM-V 4.6 1.3B

Sarvam 30B (high)

Mistral Small 3

LFM2 8B A1B

Llama 3.2 3B

Granite 4.0 H 1B

Apertus 70B Instruct

NVIDIA Nemotron Nano 12B v2 VL

Llama 3.1 Nemotron Nano 4B v1.1

Exaone 4.0 1.2B

Magistral Medium 1

Magistral Small 1

Granite 4.0 Micro

Llama 3 8B

OLMo 2 7B

Qwen3 14B

Phi-3 Mini

Ling-mini-2.0

Phi-4

Qwen3 0.6B

Exaone 4.0 1.2B

DeepSeek R1 Distill Qwen 32B

Llama 3.2 1B

Granite 3.3 8B

Apertus 8B Instruct

DeepSeek R1 Distill Qwen 14B

Gemma 3n E2B

LFM2 1.2B

Qwen3 0.6B

Qwen3.5 0.8B

Qwen3 1.7B

Phi-4 Mini

Granite 4.0 1B

Tiny Aya Global

Mistral 7B

Gemma 3 1B

DeepSeek R1 0528 Qwen3 8B

Molmo 7B-D

LFM2 2.6B

DeepSeek R1 Distill Llama 8B

Granite 4.0 H 350M

Granite 4.0 350M

DeepSeek R1 Distill Qwen 1.5B

Gemma 3 270M

List bars visualize relative performance. Plot bars use published 0–100 scores, shrink to a 1.5rem minimum, then scroll horizontally.

Scores are published snapshots and are only comparable within the same benchmark, version, metric, and evaluation configuration.